Conversational, Not a Form
Telegram-driven
A topic becomes a script, video, and voiceover through a live back-and-forth review, not a single unreviewable submission.
AI & Automation
Turns a topic into a ready-to-publish short video — through a live conversation, not a one-shot form.
A Telegram-driven pipeline that scripts, generates, narrates, and renders a short video from a single theme, with a review-and-revise checkpoint at every stage.
Business Problem
Producing a short-form video normally means writing a script, recording or sourcing a voiceover, generating or shooting visuals, and editing everything together — a multi-tool process that's slow to repeat for every new idea, and hard to revise without redoing several steps by hand.
Business Impact
Telegram-driven
A topic becomes a script, video, and voiceover through a live back-and-forth review, not a single unreviewable submission.
Each scene is individually addressable, so a revision can target one scene's video or voice instead of regenerating the whole video.
Every script, scene, and render is logged with its own status and history, so any run can be audited or resumed.
Suited For
Built for anyone who needs to turn a written idea into short-form video on a recurring basis, without operating a video editor by hand for every clip.
Goes from a theme to a finished short video through a normal chat conversation.
A repeatable script-to-video pipeline that supports revision requests instead of one-shot output.
Every stage is logged, so output can be tracked and reviewed before it goes live.
Each stage — script, video, voice, render, publish — is a swappable module, not a hard-coded chain to one provider.
If your team is manually stitching together AI scripts, generated visuals, and voiceovers, this pipeline automates that exact handoff.
Every stage — the script model, the video generator, the voice provider — can be swapped independently as better options appear.
Solution
A Telegram conversation drives the entire pipeline: a submitted theme becomes an AI-written script, which the user can approve or revise in plain text. Once approved, each scene generates its own video and voiceover, gets rendered together, and comes back for a final review — with full or partial revisions supported — before publishing.
How It Works
01
User sends a video theme to the Telegram bot.
02
AI generates a scene-by-scene script and sends it back for review.
03
User approves or requests a free-text revision of the script.
04
Each scene's video and voiceover are generated.
05
Scenes are combined into one video with synced narration.
06
User reviews the finished video and confirms it for publishing.
Architecture
Engineering Proof
Each scene has its own tracked state, so a revision can target one scene's video or voice without regenerating the entire video.
Script, video, voice, and render each run as an independent stage, so any single provider can be replaced without redesigning the pipeline.
Stages that legitimately return no data are forced to still pass through, preventing the pipeline from silently stopping mid-run.
Re-approving an already-approved step safely resumes the same run instead of starting a duplicate.
Validation
My Role
Technology
Tell me about the process, and I'll tell you what can be automated or built.